It would be accurate, if it was that short, but it is much more, even with these easily drivable FETs. Look at the gate waveform!
Read what I wrote! I wrote about miller charge only. Do the calculations yourself! Or just look at the picture of mgm2000ro! With the original devices the gate is NEVER charged up fully.
This is more than enough here.
But here is not only 100 ohm/gate, but a common 220 ohm for the global gate!
Last edited:
Maybe, but this is still insane.
The power dissipation of gate-driver doesn't depend on the gate resistor. With smaller R the current is higher, but the impulse is shorter.
I don't know if it is really that high, but gate waveform shows it is guite high really (maybe not 220 ohms), but the high side must be much stronger (judged from output waveform). B.I.G should check it!
The question is: do we still have to limit the efficiency if the output devices are replaced by much better ones?
darkfenriz said:
If we belive 2V/div for the first picture and R192=220 ohms, then turn on current can't be more than 40 mA.
I was thinking that at 200 ohms the power of the gate driving would all be lost in the resistor. (Assuming the waveform on the emitters is still square)
It might be interesting to see what happens by trying to warm up those NPN's a little bit.
Is there any/enough dead time from the modulator output?
It might be interesting to see what happens by trying to warm up those NPN's a little bit.
Is there any/enough dead time from the modulator output?
Last edited:
With 100R and a diode on each gate I'd probably try actually shorting those emitter resistors and see what happens to the supply current. Depending on what you have for a power supply you could try decreasing slowly so yo don't get any surprises. Of course you might want to make sure the gate supply isn't too high for that.
Last edited:
it looks like my driver board has R192=150 ohms and R191=100ohms R190=830ohm R189=880ohm and my driver transistors get very hot . without a fan i have yet to run them for more than a few seconds beacuse i was afraid of blowing them up.
my board does not use mpsa92 but ktc3228/kta1275
i can try the old light bold in series with the +12V but i think the driver bipolars will get fryed cause they already get hot...
my board does not use mpsa92 but ktc3228/kta1275
i can try the old light bold in series with the +12V but i think the driver bipolars will get fryed cause they already get hot...
Last edited:
Before talking about speeding up gate drive, please check the huge distance between the gate drivers and the MOSFET... Consider potential ground loops too... And don't forget the unequal current sharing that arises in such a layout when gate turn-on resisance is lowered and di/dt becomes more dependent on source lead (and PCB) inductance than on transconductance.
Slow gate drive is used for a good reason
Then look again at gate waveforms. Are you considering the effects of source lead inductance? The positive glitch indicates the beginning of conduction (charging of switching node capacitance, positive di/dt on Ls), then the negative glitch indicates that switching has been completed (capacitance charged, current drops, negative di/dt on Ls).
Throwing some low value resistors at the PCB is very easy, all newbies like to do that (I once thought it was good too).
Note that IRFP4229 does not start to conduct some current until 5.5V or so and is fully on at 6.5V
Slow gate drive is used for a good reason
Then look again at gate waveforms. Are you considering the effects of source lead inductance? The positive glitch indicates the beginning of conduction (charging of switching node capacitance, positive di/dt on Ls), then the negative glitch indicates that switching has been completed (capacitance charged, current drops, negative di/dt on Ls).
Throwing some low value resistors at the PCB is very easy, all newbies like to do that (I once thought it was good too).
Note that IRFP4229 does not start to conduct some current until 5.5V or so and is fully on at 6.5V
Last edited:
i think what Eva is telling us is that is a poor design leave it that way or else ... i will just replace the irf3205 and try again but first i need to fiind some heatsink clamps or drill the heatsink and put screws .... the hard way....
i actualy liked how it worked on my woofer ...
than it starts to smell ... that is the smell of money burning with a thick smoke too
later edit: is IRL3705ZPBF any good for replacing irf3205 ? i can not find any in stock ... so it will take a lot more time till i get some irf3205...
i actualy liked how it worked on my woofer ...
than it starts to smell ... that is the smell of money burning
with a thick smoke too later edit: is IRL3705ZPBF any good for replacing irf3205 ? i can not find any in stock ... so it will take a lot more time till i get some irf3205...
Last edited:
Replacing the IRF3205 in the push-pull SMPS by something else is not by far as troublesome as replacing the output MOSFET in a hard-switched half-bridge. Anything similar will do. For example, a few years ago I bought IRF48V and IRF2907Z in some quantity at a good price and I have been using these since then (sometimes gate resistors have to be adjusted, though).
Be aware that the market of low voltage switching MOSFET changes quickly, and parts that once were low-cost industry-standard become outdated and expensive, because they are no longer being produced in high quantity and because newer MOSFET can do the same job with less silicon. You have to look for the lowest cost recent equivalent of IRF3205 in todays market, any manufacturer will do.
For example, a quick search at digikey revealed this: Digi-Key - 785-1146-5-ND (Manufacturer - AOT460)
Note that other suppliers will probably have better prices on other devices while they may not even know that the brand A.O. Semi exists.
Concerning the schematic, I hadn't seen it in full before. It's post-filter self oscillating! With such a PCB, any excess EMI can easily disturb the modulator, making it oscillate at a much higher frequency (where something rings) and activating self destruction. But this will only happen at high output currents and near clipping, when carrier residual has the lowest slope just when switching is taking place and the highest EMI is being produced (lowest noise immunity conditions).
You should play with the value of the resistors in series with the emitters of the drivers. I suggest increasing R191 or swapping R191 and R192 for even dead time. There is not much you can do about the insane heating of the level shifter transistors, it's designed that way. Improving it requires throwing away the complete driver PCB and using another scheme. It's made to age and fail in some way that justifies throwing it away and buying a new one. It's called "Consumer Electronics" for a good reason
btw: Make it work again and you will be contributing to stop all that madness.
Be aware that the market of low voltage switching MOSFET changes quickly, and parts that once were low-cost industry-standard become outdated and expensive, because they are no longer being produced in high quantity and because newer MOSFET can do the same job with less silicon. You have to look for the lowest cost recent equivalent of IRF3205 in todays market, any manufacturer will do.
For example, a quick search at digikey revealed this: Digi-Key - 785-1146-5-ND (Manufacturer - AOT460)
Note that other suppliers will probably have better prices on other devices while they may not even know that the brand A.O. Semi exists.
Concerning the schematic, I hadn't seen it in full before. It's post-filter self oscillating! With such a PCB, any excess EMI can easily disturb the modulator, making it oscillate at a much higher frequency (where something rings) and activating self destruction. But this will only happen at high output currents and near clipping, when carrier residual has the lowest slope just when switching is taking place and the highest EMI is being produced (lowest noise immunity conditions).
You should play with the value of the resistors in series with the emitters of the drivers. I suggest increasing R191 or swapping R191 and R192 for even dead time. There is not much you can do about the insane heating of the level shifter transistors, it's designed that way. Improving it requires throwing away the complete driver PCB and using another scheme. It's made to age and fail in some way that justifies throwing it away and buying a new one. It's called "Consumer Electronics" for a good reason
btw: Make it work again and you will be contributing to stop all that madness.
Last edited:
Maybe this could be a problem, but i don't see any sign of it yet.
potential
This can be a problem, I agree.
Slow gate drive is used for a good reason
Then look again at gate waveforms. Are you considering the effects of source lead inductance?
Yes, and it can be a benefit if used properly. It can decrease di/dt without decreasing dv/dt.
You are misleaded by your experience (with very different parameters), and ignored some important datas from this case. You can see something similar in case of current turn on, but here the voltage doesn't exceed 3 V during the glitche, so it is not current turn on. Here you can see only capacitive effects, Crss couples -dv/dt from drain to gate. The -dv/dt is produced by high side turn off and current of filter choke. The observed transistor starts to conduct much later, in ZVS mode. (Meanwhile the body diode conducts, but it is seamless.)
How can you know if the glitch comes from di/dt or -dv/dt?
- from gate voltage. di/dt makes voltage higher then treshold, -dv/dt makes it lower.
- compare gate waveform of BIG and mgm2000! With IRFP4229 we see lower glitche then with IRFP264, despite the higher proposed di/dt. Contrary to this, the lower Crss capacitance of IRFP4229 explains perfectly the difference assuming that the glitche is caused by -dv/dt.
- from general experience. Here we can see idle state, currents are very low, but gate driver impedance is extremely high, so capacitive effects can arise much more easily.
- etc...
Measuring with output current may show some issue related to di/dt and PCB layout.
Last edited:
so if i play with R192/191 i can increase and decrease dead time ?
i will try one more time with an external smps that has current protect and about +/-95V and if it blows another set of irfp4229 i`ll try to mess with resistors and if not i`ll put the original irfp264 in ... if those die then i guess i have a big 12V to +/-90V psu and 12Kg heatsink hmmm 2Kw class AB anyone ?
i still think that the output fets did not touch the heatsink firmley so they overheated and blew...
thank you all for the replies ! i`ll keep you posted if i get it working right
i will try one more time with an external smps that has current protect and about +/-95V and if it blows another set of irfp4229 i`ll try to mess with resistors and if not i`ll put the original irfp264 in ... if those die then i guess i have a big 12V to +/-90V psu and 12Kg heatsink hmmm 2Kw class AB anyone ?
i still think that the output fets did not touch the heatsink firmley so they overheated and blew...
thank you all for the replies ! i`ll keep you posted if i get it working right
Am I reading that right that the driver supplies are ground and + rail, and that if for some reason output oscillation stopped or slowed too much full output supply voltage would be applied to the gates through those resistors until breakdown occurred? That can't be right. The gate voltage in the first posted oscillogram looks like it's almost reaching driver supply. 8-10V?
Last edited:
Pafi: Ok, assuming that there would be minimum or no overlap between high and low sides, the glitches would be a mix of dv/dt and di/dt (any dv/dt at Cds will cause di/dt at Ls too, and Ls is high due to the layout and the TO-247)
What makes you think that the low flat part of gate waveform is actually 0V? The driver dictates that the gates can't go to ground (they probably go to +1V or so). I think the positive glitch is slightly over 4V and the MOSFET are hot. I see a slight increase in rising Vgs slope just before the negative part of the glitch that can't have any other meaning than "start of conduction".
The pure dv/dt effect that you explain produces a different glitch with a more rounded transition to the negative part on Vgs. There isn't anything wrong with that, it's optimum "no-dead-time" operation (only on low side), but it requires a circuit with more stable and accurate timing.
Switching should be fine if high side timing is adjusted to match low side. Low side seems ok. All output MOSFET will idle moderately and evenly warm with no heatsink if this is done properly.
What makes you think that the low flat part of gate waveform is actually 0V? The driver dictates that the gates can't go to ground (they probably go to +1V or so). I think the positive glitch is slightly over 4V and the MOSFET are hot. I see a slight increase in rising Vgs slope just before the negative part of the glitch that can't have any other meaning than "start of conduction".
The pure dv/dt effect that you explain produces a different glitch with a more rounded transition to the negative part on Vgs. There isn't anything wrong with that, it's optimum "no-dead-time" operation (only on low side), but it requires a circuit with more stable and accurate timing.
Switching should be fine if high side timing is adjusted to match low side. Low side seems ok. All output MOSFET will idle moderately and evenly warm with no heatsink if this is done properly.
Last edited:
Just wanted to mention, after seeing the full amp drawing, the C4793's (NPN drivers) have lots of voltage across them at all times (you say 90V rails?) and though source current steps technically operate in linear mode and probably dissipate plenty between the pair so that you actually need both, contrary to what I assumed earlier.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.